From sounds to lexical representations

Average English-native 20-year-old knows ~42,000 words: mental lexicon

Lexical development:

• Vocabulary size (Braginsky et al. 2019)
• Semantic and phonological associations (Fourtassi, Bian, and Frank 2020)
• Efficiency in lexical access (Marchman and Fernald 2008)

From sounds to lexical representations

From sounds to lexical representations

Vocabulary growth

Vocabulary size grows non-linearly during the second year of life

Figure 1: Vocabulary size norms for 51,800 monolingual children learning 35 distinct languages (wordbank)

Early dynamics of lexical processing

In children (Chow, Davies, and Plunkett 2017)

• Semantic priming (Styles and Plunkett 2009; Arias-Trejo and Plunkett 2009)
• Phonological priming (Mani and Plunkett 2010; Avila-Varela, Arias-Trejo, and Mani 2021)

Challenges of a bilingual environment

• Increased complexity in linguistic context (learning two codes)
• Reduced linguistic input (split into two languages)
• Increased referential ambiguity

Challenges of a bilingual environment

Bilinguals acquire words at similar rates as monolinguals (Hoff et al. 2012)

A cognate facilitation in lexical acquisition?

Lexically closer languages ➡️ Larger vocabulary size (Floccia et al. 2018)

English-Dutch > English-Mandarin

Cognates acquired earlier than non-cognates (Mitchell, Tsui, and Byers-Heinlein 2023; Bosch and Ramon-Casas 2014)

A cognate facilitation in lexical acquisition?

Figure 2: Pairwise lexical similarity (average Levensthein similarity across translations).

A cognate facilitation in lexical acquisition?

Introduction

• Bilinguals keep up with monolinguals in vocabulary growth
• The presence of cognates seems to facilitate vocabulary acquisition
• Language co-activation has been suggested to underlie this phenomenon
• Mechanisms supporting the role of co-activation in cognate acquisition are unclear

Language non-selectivity

Language non-selectivity

The AMBLA model

\[ \begin{aligned} \definecolor{myred}{RGB}{ 168, 0, 53 } \definecolor{myblue}{RGB}{ 0, 64, 168 } \definecolor{mygreen}{RGB}{0, 168, 87} \definecolor{grey}{RGB}{128, 128, 128} \textbf{For participant } &i \textbf{ and word-form } j \text{ (translation of } j'): \\ {\color{mygreen}\text{Age of Acquisition}_{ij}} &= \{\text{Age}_i \mid {\color{myred}\text{Learning instances}_{ij}} = {\color{myblue}\text{Threshold}} \}\\ \color{myred}{\text{Learning instances}_{ij}} &= \text{Age}_i \cdot \text{Freq}_j \\ \textbf{where:} \\ {\color{myblue}\text{Threshold}} &= 300 \\ \text{Freq}_j &\sim \text{Poisson}(\lambda = 50) \end{aligned} \]

Extended to the bilingual case:

Accumulation of learning instances, a function of Exposure and Frequency.

\[ \begin{aligned} \definecolor{myred}{RGB}{ 168, 0, 53 } \definecolor{myblue}{RGB}{ 0, 64, 168 } \definecolor{mygreen}{RGB}{0, 168, 87} \definecolor{myorange}{RGB}{ 235, 127, 26 } \textbf{For participant } &i \textbf{ and word-form } j \text{ (translation of } j'): \\ \text{Age of Acquisition}_{ij} &= \{\text{Age}_i \mid \text{Learning instances}_{ij} = \text{Threshold} \}\\ \text{Learning instances}_{ij} &= \text{Age}_i \cdot \text{Freq}_j \cdot {\color{myred}\text{Exposure}_{ij}}\\ \textbf{where:} \\ \text{Threshold} &= 300 \\ \text{Freq}_j &\sim \text{Poisson}(\lambda = 50) \end{aligned} \]

Implementing a cognateness facilitation mechanism:

\[ \begin{aligned} \definecolor{myred}{RGB}{ 168, 0, 53 } \definecolor{myblue}{RGB}{ 0, 64, 168 } \definecolor{mygreen}{RGB}{0, 168, 87} \definecolor{myorange}{RGB}{ 235, 127, 26 } \textbf{For participant } &i \textbf{ and word-form } j \text{ (translation of } j'): \\ \text{Age of Acquisition}_{ij} &= \{\text{Age}_i \mid \text{Learning instances}_{ij} = \text{Threshold} \}\\ \text{Learning instances}_{ij} &= \text{Age}_i \cdot \text{Freq}_j \cdot \text{Exposure}_{ij} \cdot {\color{myred}\text{Cognateness}_{j}}\\ \textbf{where:} \\ \text{Threshold} &= 300 \\ \text{Freq}_j &\sim \text{Poisson}(\lambda = 50) \\ {\color{myred}\text{Cognateness}}&{\color{myred}=\text{Levenshtein}(j, j')} \end{aligned} \]

Methods

• Barcelona Vocabulary Questionnaire (BVQ): vocabulary checklist, language exposure questionnaire, demographic survey
  • Online, inspired by Oxford CDI (Hamilton, Plunkett, and Schafer 2000)
  • 793 Catalan words, 797 Spanish words, divided into 4 sublists
• N = 366 children, 12-32 months (M = 22.23, SD = 4.88, Range = 12.06–31.93)
• Exposed to Catalan and Spanish (+ up to 10% to a 3rd languages)

Methods

• 302 translation equivalents
• 234 questionnaire administrations
• 138,078 item responses (No, Understands, Understands and Says)

Ordinal, multilevel regression model

response ~ age * exposure * cognateness + length + (...|child) + (...|te) 

\[ \begin{aligned} \text{Exposure}_{ij} &= \text{Frequency}_j \times \text{Language degree of exposure}_{ij} \\ \text{Cognateness}_{j} &= \text{Levenshtein}(j, j') \end{aligned} \]

Bayesian: \[ \text{Posterior} \propto \text{Prior} \times \text{Likelihood} \]

Results: fixed regression coefficients

Figure 4: Posterior distribution of fixed regression coefficients

Results: marginal posterior predictions

Figure 5: Posterior marginal effects

Discussion

• Earlier acquisition for cognates vs. non-cognates
• Cognate facilitation moderated by exposure

Only words from the lower exposure benefit from cognateness Parallel to language dominance effects in adults?

• Cognateness as a candidate mechanism underlying Floccia et al. (2018)’s results
• Cross-language facilitation via co-activation of phonologically similar translation equivalents

Language non-seletivity in the initial lexicon

Cross-language phonological priming:

Von Holzen and Mani (2012): German-English (N = 20, 21-43 months), priming through translation paradigm

Language non-seletivity in the initial lexicon

Mani and Plunkett (2010): English monolinguals, implicit priming

Hypotheses

Experiment 1

N = 112 children (15 longitudinal)

Average age 26.36 months (SD = 4.01, Range = 20.03–32.5)

English monolinguals, Oxford (United Kindgom)

Figure 6: Participant receptive vocabulary sizes across ages and language profiles.

Design

Results

Experiment 2

N = 112 children (15 longitudinal)

Average age 26.36 months (SD = 4.01, Range = 20.03–32.5)

English monolinguals, Oxford (United Kindgom)

Figure 7: Participant receptive vocabulary sizes across ages and language profiles.

Results (monolinguals)

Results (bilinguals)

Discussion